(T-112) Early sampling improves voriconazole target attainment in pediatric patients dosed with model-informed precision dosing tools: a simulation study

Dominic Tong – InsightRX; Laura Bio – Stanford Medicine Children's Health – Lucile Packard Children's Hospital Stanford; Kevin Downes – Children's Hospital of Philadelphia; Anna Sharova – Children's Hospital of Philadelphia; Jasmine Hughes – InsightRX; Maria-Stephanie Hughes – InsightRX

Objectives: Model-informed precision dosing (MIPD) holds promise for optimizing voriconazole therapy in pediatric populations by enabling earlier, more accurate pharmacokinetic (PK) assessments and thus individualized dosing. Dosing is challenging due to its nonlinear PK and metabolism via cytochrome P450 (CYP) enzymes, particularly CYP2C19 [1, 2]. Therapeutic drug monitoring (TDM) addresses these issues; however, without MIPD, only steady-state samples are interpretable, delaying dose adjustment and target attainment. Our previous work showed that a priori predictions based on population pharmacokinetic (popPK) models and patient covariates exhibit substantial bias and imprecision [3]. Given the risks associated with both subtherapeutic exposure (treatment failure or breakthrough invasive fungal infection) and supratherapeutic (toxicity), this study evaluates how earlier sampling can improve target trough attainment in pediatric patients using a simulated MIPD trial.

Methods: Two published popPK models (Friberg [1] and Muto [2]) were used to simulate patients. These models previously showed best predictive performance in a retrospective 160 patient cohort from two sites [3]. We used mipdtrial [4], an R package for evaluating iterative MIPD dosing strategies, to simulate 100 virtual pediatric patients based on age and weight distributions from real data. Individual PK variability and sampling error were simulated from model parameters. Each simulated patient was initially dosed with 6 mg/kg every 12 hours for one day, then 4 mg/kg every 12 hours. For patients younger than 14 years old and weighing less than 40kg, the dose amounts were 9 mg/kg and 8 mg/kg, respectively. PK parameters were estimated using maximum a posteriori (MAP) Bayesian fitting with the other model, mimicking a real-world scenario where the estimating model is not a perfect descriptor of patient PK. We then evaluated the impact of earlier sample timing (days 2–5) with varying sample turnaround time (1–3 days), both with and without an added steady-state sample four days later. The primary outcome was evaluating target trough attainment (1–6 mg/L) on each day for 14 days after initiation of therapy.

Results: When simulating with the Muto model and estimating with the Friberg model, early sampling on day 2, with a dose adjustment one day later, improved day 3 target attainment compared to sampling on day 5 from 73% to 87% (+14%). However, this led to supratherapeutic exposures in 40% of patients by day 14 compared to only 8% with a single day 5 sample. Adding a second sample on day 6 maintained early target attainment while reducing day 14 overexposure to 6%. These trends were consistent when model roles were reversed, though the Friberg model’s time-dependent clearance effect may have contributed to overestimation of elimination with early-only sampling.

Conclusions: Early sampling and dose adjustment of voriconazole, especially when paired with a subsequent steady-state sample, improves early target attainment in pediatric patients without increasing overexposure risk. These findings support incorporating early sampling into routine MIPD workflows to optimize voriconazole therapy and mitigate risks of therapeutic failure and toxicity.

Citations: [1] Friberg et al, AAC, 2012
[2] Muto et al, AAC, 2015
[3] Hughes MA et al, IDWeek 2025 abstract (under submission)
[4] insightrx.github.io/mipdtrial/

Keywords: Model-informed Precision Dosing, Pediatric pharmacokinetics, antifungal