Evaluation of a Bayesian Dosing Calculator for Vancomycin in Pediatric Patients with Augmented Renal Clearance

Abdullah Alsultan^1*Muneera R Aljelaify²Huda Alshahrani³Noura M Alajmi⁴Ghadeer Alfuhaydi⁵Saeed Alqahtani¹Ali Somily¹Mashal Almutairi¹Manal Aboelkheir^6*

• ¹King Saud University, Riyadh, Saudi Arabia
• ²King Saud University Medical City, Riyadh, Saudi Arabia
• ³King Khalid University, Abha, Saudi Arabia
• ⁴King Abdulaziz Medical City - Jeddah, Jeddah, Saudi Arabia
• ⁵Qassim University, Buraydah, Saudi Arabia
• ⁶Misr International University, Cairo, Egypt

Background: Augmented renal clearance (ARC) is increasingly recognized among pediatric oncology and intensive care patients. It can result in subtherapeutic concentrations of renally eliminated drugs like vancomycin. Bayesian dosing tools are recommended to individualize therapy, yet their performance in pediatric ARC remains underexplored. This study evaluated the predictive accuracy of the freely available Bayesian dosing calculator, NextDose, for vancomycin in pediatric patients with and without ARC. NextDose is a web-based application built on a large-population pharmacokinetic model encompassing neonates to adults with varying renal function. Methods: A retrospective observational study included pediatric patients (1–12 years) who received vancomycin and had at least one steady-state serum concentration. ARC was defined as estimated glomerular filtration rate (eGFR) >130 mL/min/1.73m². Predictive performance was assessed using relative median prediction error (rMPE, bias) and relative median absolute prediction error (rMAPE, precision; lower values indicate higher precision). Both a priori (using only clinical/demographic data) and a posteriori (drug-level informed) predictions were evaluated. Results: A total of 112 pediatric patients were included, of whom 47 (42%) met ARC criteria. The mean age was 5.9 ± 3.4 years; 10.7% were younger than 2 years, 57.2% were aged 2–7 years, and 32.1% were 7–12 years old. A priori predictions showed high bias (rMPE 27%) and moderate precision (rMAPE 31%), with no significant differences between the ARC and non-ARC groups. In contrast, a posteriori predictions demonstrated marked improvement (rMPE −3.9%, rMAPE 13.5%), with 86% of predictions meeting the <30% prediction-error threshold. Patients with ARC exhibited superior predictive accuracy than non-ARC counterparts (rMAPE 12% vs. 17.5%, p=0.03). Conclusion: NextDose overestimated vancomycin concentrations in a priori predictions, suggesting it may not be suitable for initial dose calculations. Incorporating one or two measured concentrations significantly improved predictive accuracy, particularly in patients with ARC, supporting its use alongside therapeutic drug monitoring to personalize vancomycin monitoring.