High VWF:Act and accumulation of FVIII were observed after perioperative FVIII-based replacement therapy in patients with VWD, both underlining the necessity of personalization of dosing regimens to optimize perioperative treatment.

According to GlaxoSmithKline's Clinical Trial Register, data from the GlaxoSmithKline studies LAM100034 and LEP103944, corresponding to ClinicalTrials.gov identifiers NCT00113165 and NCT00264615, used in this work, have been used in previous publications (doi: https://doi.org/10.1212/01.wnl.0000277698.33743.8b , https://doi.org/10.1111/j.1528-1167.2007.01274.x ).

No abstract

Nonlinear mixed effect (NLME) models are the gold standard for the analysis of patient response following drug exposure. However, these types of models are complex and time‐consuming to develop. There is great interest in the adoption of machine‐learning methods, but most implementations cannot be reliably extrapolated to treatment strategies outside of the training data. In order to solve this problem, we propose the deep compartment model (DCM), a combination of neural networks and ordinary differential equations. Using simulated datasets of different sizes, we show that our model remains accurate when training on small data sets. Furthermore, using a real‐world data set of patients with hemophilia A receiving factor VIII concentrate while undergoing surgery, we show that our model more accurately predicts a priori drug concentrations compared to a previous NLME model. In addition, we show that our model correctly describes the changing drug concentration over time. By adopting pharmacokinetic principles, the DCM allows for simulation of different treatment strategies and enables therapeutic drug monitoring.

In population pharmacokinetic (PK) models, interindividual variability is explained by implementation of covariates in the model. The widely used forward stepwise selection method is sensitive to bias, which may lead to an incorrect inclusion of covariates. Alternatives, such as the full fixed effects model, reduce this bias but are dependent on the chosen implementation of each covariate. As the correct functional forms are unknown, this may still lead to an inaccurate selection of covariates. Machine learning (ML) techniques can potentially be used to learn the optimal functional forms for implementing covariates directly from data. A recent study suggested that using ML resulted in an improved selection of influential covariates. However, how do we select the appropriate functional form for including these covariates? In this work, we use SHapley Additive exPlanations (SHAP) to infer the relationship between covariates and PK parameters from ML models. As a case-study, we use data from 119 patients with hemophilia A receiving clotting factor VIII concentrate peri-operatively. We fit both a random forest and a XGBoost model to predict empirical Bayes estimated clearance and central volume from a base nonlinear mixed effects model. Next, we show that SHAP reveals covariate relationships which match previous findings. In addition, we can reveal subtle effects arising from combinations of covariates difficult to obtain using other methods of covariate analysis. We conclude that the proposed method can be used to extend ML-based covariate selection, and holds potential as a complete full model alternative to classical covariate analyses. Study Highlights WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?Covariate selection in pharmacokinetic (PK) modeling is a complex process and is sensitive to bias. Machine-learning (ML) techniques might help to simplify and potentially improve this process, but are difficult to interpret as is.

Objective Most von Willebrand disease (VWD) patients can be treated with desmopressin during bleeding or surgery. Large interpatient variability is observed in von Willebrand factor (VWF) activity levels after desmopressin administration. The aim of this study was to develop a pharmacokinetic (PK) model to describe, quantify, and explain this variability. Methods Patients with either VWD or low VWF, receiving an intravenous desmopressin test dose of 0.3 µg kg−1, were included. A PK model was derived on the basis of the individual time profiles of VWF activity. Since no VWF was administered, the VWF dose was arbitrarily set to unity. Interpatient variability in bioavailability (F), volume of distribution (V), and clearance (Cl) was estimated. Results The PK model was developed using 951 VWF activity level measurements from 207 patients diagnosed with a VWD type. Median age was 28 years (range: 5–76), median predose VWF activity was 0.37 IU/mL (range: 0.06–1.13), and median VWF activity response at peak level was 0.64 IU/mL (range: 0.04–4.04). The observed PK profiles were best described using a one-compartment model with allometric scaling. While F increased with age, Cl was dependent on VWD type and sex. Inclusion resulted in a drop in interpatient variability in F and Cl of 81.7 to 60.5% and 92.8 to 76.5%, respectively. Conclusion A PK model was developed, describing VWF activity versus time profile after desmopressin administration in patients with VWD or low VWF. Interpatient variability in response was quantified and partially explained. This model is a starting point toward more accurate prediction of desmopressin dosing effects in VWD.

Introduction Many patients with von Willebrand disease (VWD) are treated on demand with von Willebrand factor and factor VIII (FVIII) containing concentrates present with VWF and/or FVIII plasma levels outside set target levels. This carries a risk for bleeding and potentially for thrombosis. Development of a population pharmacokinetic (PK) model based on FVIII levels is a first step to more accurate on‐demand perioperative dosing of this concentrate. Methods Patients with VWD undergoing surgery in Academic Haemophilia Treatment Centers in the Netherlands between 2000 and 2018 treated with a FVIII/VWF plasma‐derived concentrate (Haemate® P/Humate P®) were included in this study. Population PK modeling was based on measured FVIII levels using nonlinear mixed‐effects modeling (NONMEM). Results The population PK model was developed using 684 plasma FVIII measurements of 97 VWD patients undergoing 141 surgeries. Subsequently, the model was externally validated and reestimated with independent clinical data from 20 additional patients undergoing 31 surgeries and 208 plasma measurements of FVIII. The observed PK profiles were best described using a one‐compartment model. Typical values for volume of distribution and clearance were 3.28 L/70 kg and 0.037 L/h/70 kg. Increased VWF activity, decreased physical status according to American Society of Anesthesiologists (ASA) classification (ASA class >2), and increased duration of surgery were associated with decreased FVIII clearance. Conclusion This population PK model derived from real world data adequately describes FVIII levels following perioperative administration of the FVIII/VWF plasma‐derived concentrate (Haemate® P/Humate P®) and will help to facilitate future dosing in VWD patients.

Recent studies have reported that patients with von Willebrand disease treated perioperatively with a von Willebrand factor (VWF)/factor VIII (FVIII) concentrate with a ratio of 2.4:1 (Humate P/Haemate P) often present with VWF and/or FVIII levels outside of prespecified target levels necessary to prevent bleeding. Pharmacokinetic (PK)-guided dosing may resolve this problem. As clinical guidelines increasingly recommend aiming for certain target levels of both VWF and FVIII, application of an integrated population PK model describing both VWF activity (VWF:Act) and FVIII levels may improve dosing and quality of care. In total, 695 VWF:Act and 894 FVIII level measurements from 118 patients (174 surgeries) who were treated perioperatively with the VWF/FVIII concentrate were used to develop this population PK model using nonlinear mixed-effects modeling. VWF:Act and FVIII levels were analyzed simultaneously using a turnover model. The protective effect of VWF:Act on FVIII clearance was described with an inhibitory maximum effect function. An average perioperative VWF:Act level of 1.23 IU/mL decreased FVIII clearance from 460 mL/h to 264 mL/h, and increased FVIII half-life from 6.6 to 11.4 hours. Clearly, in the presence of VWF, FVIII clearance decreased with a concomitant increase of FVIII half-life, clarifying the higher FVIII levels observed after repetitive dosing with this concentrate. VWF:Act and FVIII levels during perioperative treatment were described adequately by this newly developed integrated population PK model. Clinical application of this model may facilitate more accurate targeting of VWF:Act and FVIII levels during perioperative treatment with this specific VWF/FVIII concentrate (Humate P/Haemate P).