Towards Model-Informed Precision Dosing of Voriconazole: Challenging Published Voriconazole Nonlinear Mixed-Effects Models with Real-World Clinical Data

Abstract

Background and objectives: Model-informed precision dosing (MIPD) frequently uses nonlinear mixed-effects (NLME) models to predict and optimize therapy outcomes based on patient characteristics and therapeutic drug monitoring data. MIPD is indicated for compounds with narrow therapeutic range and complex pharmacokinetics (PK), such as voriconazole, a broad-spectrum antifungal drug for prevention and treatment of invasive fungal infections. To provide guidance and recommendations for evidence-based application of MIPD for voriconazole, this work aimed to (i) externally evaluate and compare the predictive performance of a published so-called 'hybrid' model for MIPD (an aggregate model comprising features and prior information from six previously published NLME models) versus two 'standard' NLME models of voriconazole, and (ii) investigate strategies and illustrate the clinical impact of Bayesian forecasting for voriconazole.

Methods: A workflow for external evaluation and application of MIPD for voriconazole was implemented. Published voriconazole NLME models were externally evaluated using a comprehensive in-house clinical database comprising nine voriconazole studies and prediction-/simulation-based diagnostics. The NLME models were applied using different Bayesian forecasting strategies to assess the influence of prior observations on model predictivity.

Results: The overall best predictive performance was obtained using the aggregate model. However, all NLME models showed only modest predictive performance, suggesting that (i) important PK processes were not sufficiently implemented in the structural submodels, (ii) sources of interindividual variability were not entirely captured, and (iii) interoccasion variability was not adequately accounted for. Predictive performance substantially improved by including the most recent voriconazole observations in MIPD.

Conclusion: Our results highlight the potential clinical impact of MIPD for voriconazole and indicate the need for a comprehensive (pre-)clinical database as basis for model development and careful external model evaluation for compounds with complex PK before their successful use in MIPD.

Fig. 1

Schematic workflow of the external model evaluation framework. MIPD model-informed precision dosing, NLME nonlinear mixed-effects

Fig. 2

Identified nonlinear mixed-effects pharmacokinetic models for voriconazole (aggregate model [35], M1 [6] and M2 [13]) (a, upper panel) and deterministic simulations for a 60-year-old, 67 kg typical patient (lean body weight 54 kg) and CYP2C19*1/*1 genotype (normal metabolizer, wild type), without co-medication using all models (lower panels). Typical voriconazole concentration-time profiles in plasma (b, top lower left panel), and total voriconazole clearance over time (c, bottom lower left panel) after administration of 400 mg voriconazole intravenously over 2 h (400 mg IV, SD1) and 400 mg per orally (400 mg PO, SD2). Typical concentration-time profiles obtained after administration of different multiple IV and PO dosing regimens with (2 × 6 mg/kg IV q12h, then 4 mg/kg IV q12h, MD1, 2 × 400 mg IV q12h, then 200 mg PO q 12h, MD3)/without (4 mg/kg IV q12h, MD2, 200 mg PO q12h, MD4) loading doses (d, lower right panels). Solid red line: aggregate model, Dotted blue line: M1, Dashed green line: M2, Dotted black lines (d): lower (1.0 mg/L) and upper bound (6 mg/L) of C_min target range for voriconazole. *Interindividual variability included on parameter, ⁺Interoccasion variability included on parameter, ^#Parameter fixed to literature value. AGE covariate effect of age, ALP covariate effect of alkaline phosphatase, CHOL covariate effect of severe hepatic cholestasis, CL linear clearance, CL_NL time-dependent nonlinear clearance [V_max/(K_M + C_P), with V_max maximum elimination rate, CLT total clearance, C_min minimum plasma concentration, C_P plasma concentration, K_M Michaelis-Menten constant, EIND covariate effect of enzyme inducer coadministration, EINH covariate effect of enzyme inhibitor coadministration, F oral bioavailability, IV intravenous, k_A absorption rate constant, LBW covariate effect of lean body weight (allometric scaling), PM covariate effect of CYP2C19 genotype-predicted phenotype “poor metabolizer”, PO per oral, Q intercompartmental clearance, RIF covariate effect of rifampicin coadministration, WT covariate effect of total body weight (allometric scaling)

Fig. 3

Population-predicted versus observed voriconazole concentrations using external evaluation dataset stratified by nonlinear mixed-effects pharmacokinetic model of voriconazole. Dotted black line: line of identity, Dashed red line: loess-regression line

Fig. 4

Prediction-corrected visual predictive checks (n = 1000 simulations) using the ‘predictive performance evaluation dataset’ (n_samples = 329) comprising data from 31 patients after administration of 400 mg voriconazole (a, loading dose, upper panels) and after administration of 200 mg voriconazole at steady state (b, maintenance dose, lower panels) stratified by nonlinear mixed-effects pharmacokinetic model of voriconazole. Black circles: observed voriconazole concentrations, Lines: 5th, 95th percentile (dashed), 50th percentile (solid) of observed (red) and simulated (black) data, Blue shaded area: 95% CIs around 5th, 95th percentile of simulated data, Red shaded area: 95% CIs around 50th percentile of simulated data, Red stars: percentiles of observed data outside of 95% CIs around percentiles of simulated data

Fig. 5

Relative mean prediction error (a, c) and relative root mean square error (b, d) of the predicted versus the observed voriconazole concentrations following the 6th occasion stratified by nonlinear mixed-effects pharmacokinetic model of voriconazole. The individual predicted voriconazole concentrations of the ‘most recent’ (i.e., 6th) occasion were forecasted by covariate information only (a priori prediction) and using the first one, two, three, four, and five prior observations (strategies a1–a6, left panels) or using different combinations of up to three prior observations (strategies b1–b7, right panels)

Fig. 6

A schematic model-informed precision dosing workflow for voriconazole (a, left panel) and simulation to illustrate the clinical impact of Bayesian forecasting (b, right panel). Red line typical: voriconazole concentration-time profile in plasma of a reference patient (based on the characteristics of the ‘predictive performance evaluation dataset’) receiving the standard dosing regimen (MD3) using the aggregate nonlinear mixed-effects pharmacokinetic model of voriconazole, Black cross: C_min sample taken at Day 3, Blue line: patient-individual concentration-time profile assimilating C_min sample into Bayesian forecast, Green line: updated patient-individual concentration-time profile based on C_min sample and optimized dose of 350 mg, Grey shaded area: indicating future simulation time, Dotted black line: lower bound of C_min target range for voriconazole. C_min minimum plasma concentration, C(t) profile concentration-time profile, MIPD model-informed precision dosing, NLME nonlinear mixed-effects, PK pharmacokinetic(s), PO per oral