Author’s Reply to Helsby and Hannam: ‘Understanding Voriconazole Metabolism: A Middle-Out Physiologically-Based Pharmacokinetic Modelling Framework Integrating In Vitro and Clinical Insights’

We thank the authors of the letter to the editor, Helsby and Hannam [1], for their constructive comments on our recent publication describing a middle-out PBPK modelling framework for voriconazole and its metabolites [2]. The authors highlight important points, particularly regarding the distinction between 4-hydroxyvoriconazole (4-OHVRC) and hydroxyvoriconazole (OHVRC), which indeed represent different chemical entities. Their observations provide a valuable opportunity to clarify our modelling assumptions and rationale.

As correctly highlighted, the enzyme(s) responsible for the formation of OHVRC, defined as hydroxylation on the fluoropyrimidine ring, remain(s) unidentified. This knowledge gap was explicitly acknowledged in our manuscript (“Other metabolites, such as hydroxyvoriconazole (OHVRC) and dihydroxyvoriconazole, are primarily detected in urine along with their conjugates [3‐5]. Nevertheless, their formation pathway remains elusive, highlighting a significant knowledge gap [4, 6, 7]”) [1]. OHVRC was not the primary focus of our modelling work; however, its potent in vitro inhibitory potential against CYP3A4, nearly comparable to that of the parent compound voriconazole (VRC), as demonstrated by Schulz et al. [7], and thus impacting overall the pharmacokinetics of the parent compound VRC, provided strong justification for its inclusion in the PBPK model.

In the absence of available in vitro kinetic data describing the formation of OHVRC, we assumed a not-yet-specified enzymatic elimination process in the model parameterised by CYP3A4 kinetic parameters published by Murayama et al. [8] for 4-OHVRC as a surrogate for mechanistic modelling. We would like to clarify that this decision was not based on the assumption that the same enzyme necessarily catalyses both hydroxylation reactions but rather that the kinetic values could serve as a pragmatic placeholder for enabling exploratory simulations. While both hydroxy metabolites share the same molar mass (365.32 g/mol) and show similar physicochemical properties (e.g. logP: 2.21 for OHVRC versus 1.79 for 4-OHVRC, and basic pKa ~2.0), we fully acknowledge that such properties alone do not imply similar enzyme binding or metabolic behaviour, particularly for CYP-mediated reactions.

However, our approach was supported by the following comparative analysis of formation kinetics of OHVRC and 4-OHVRC. Using the reported in vivo plasma metabolic clearance of VRC to OHVRC (20.2 ± 9.95 mL/min) from Geist et al. [6], and assuming a well-stirred liver model (equation 1), a fraction unbound in plasma of 0.5 and hepatic blood flow of 1750 mL/min, we back-calculated a hepatic intrinsic clearance (CL_int) of 41 ± 20 mL/min, corresponding to a specific hepatic clearance of ~ 0.03 ± 0.01 1/min in PK-Sim^®. By comparison, the in vitro-derived CYP3A4 kinetics (K_m = 11 µM, K_cat = 0.10 1/min) implemented in the model yielded as product a specific hepatic clearance value of 0.01 1/min, resulting in a predicted OHVRC area under the curve (AUC) of 1.99 mg/h/L following a 400 mg intravenous (IV) dose. This AUC value closely matched the predicted AUC value of 2.12 mg/h/L via the in vivo-derived process. The kinetic formation parameters included for OHVRC formation were thus, even though they were based on in vitro formation of 4-OH-VRC, similar to the ones observed in in vivo formation of OHVRC.

Equation 1: Derivation of the intrinsic clearance (CL_int) parameter from the well-stirred liver model

CL_H hepatic plasma clearance, f_u,p fraction unbound in plasma, Q_H hepatic blood flow.

Given these considerations, our use of CYP3A4 as a surrogate for OHVRC formation was not intended to imply a definitive enzymatic pathway. Rather, it enabled mechanistic implementation of a clinically observed metabolite with known inhibition relevance in the absence of definitive formation data. OHVRC has been consistently identified in both plasma and urine of patients, as reported by Scholz et al. [4] and Geist et al. [6]. In contrast, 4-OHVRC has not been confirmed in clinical samples, and Geist et al. explicitly stated that hydroxylation of the methyl group (i.e. formation of 4-OHVRC) was not detected in their patient population. These findings further support our decision to focus on OHVRC in the clinical context. As the authors rightly point out, our PBPK framework was designed to enable iterative hypothesis refinement, and this modelling choice reflects exactly that scientific intention. We agree that future work should explore other enzymatic pathways for OHVRC formation, ideally based on new experimental identification of its metabolic origin. This is especially important when expanding the model to oral administration of VRC, as CYP3A4 is also expressed in the intestine and its (auto-)inhibition can lead to changes in bioavailability (F). Therefore, incorrectly addressing OHVRC formation to CYP3A4 could lead to biased estimates of this inhibition and subsequent impact on F. A simple solution would be to avoid making assumptions about the enzyme involved in the pathway and using PO (by mouth) and IV data to re-estimate the OHVRC formation and CYP3A4 inhibition parameters. Until such in vitro and in vivo data become available, our approach represents a plausible and justifiable solution within the limits of current knowledge, particularly in light of OHVRC’s measurable impact on CYP3A4 activity and systemic exposure.

We would also like to clarify one more point regarding the interpretation of our second hypothesis, as we may have misunderstood the letter: both hypotheses incorporated the inhibitory effects of NO and OHVRC on VRC clearance. The key distinction in hypothesis 2 was the inclusion of auto-inhibition of NO clearance to reflect potential NO concentration-dependent elimination behaviour.

VRC remains a cornerstone therapy for invasive fungal infections, yet its clinical utility is hampered by significant interindividual variability in exposure, leading to risks of subtherapeutic failure or supratherapeutic toxicity. A robust mechanistic understanding of its pharmacokinetics, including metabolite-mediated inhibition and nonlinear elimination pathways, is critical to optimise dosing strategies and improve patient outcomes. Our model aims to address these challenges by bridging gaps in current knowledge, ultimately supporting safer and more effective use of this antifungal. In this regard, we are grateful to the authors of the letter for their contribution and hope this clarification supports ongoing research into the complex pharmacokinetics of VRC and its metabolites.