Van Dong Nguyen1,2,5, Cheng Ji1,2, Vlad Alexandru Rosu1,2, France Dion1,3, Marie-Ève Legris1,3, Amélie Marsot1,4,5
1Faculty of Pharmacy, Université de Montréal, 2Pharmacy Department, McGill University Healthcare Centre, 3Département de pharmacie, Hôpital Charles- Le Moyne, 4Centre de recherche du Centre hospitalier universitaire Ste-Justine, 5 Laboratoire de Suivi Thérapeutique Pharmacologique et Pharmacocinétique
Introduction: Intravenous (IV) vancomycin remains a cornerstone of antimicrobial therapy for patients undergoing chronic hemodialysis (HD), given the increased risk of hospital-acquired infection caused by methicillin resistant organisms. Although several population pharmacokinetic (popPK) models of vancomycin have been developed for this purpose, their predictive performance may be inadequate across all populations, as shown in our recent external evaluation of popPK models of vancomycin in hemodialysis [1]. In our multicenter cohort, most models exhibited a general tendency to overpredict in population predictions (PRED). In our previous work, we compared alternative methods for modeling vancomycin pharmacokinetics in intermittent HD: defining a single clearance (CL) parameter for patients on HD versus implementing separate parameters to distinguish between intradialytic and interdialytic CL. In this current analysis, we aim to explore other clinical covariates in patients receiving intermittent HD that may inform future adaptations or the development of new popPK vancomycin models for this population. Objectives: •Compare predictive performance of existing popPK models according to clinical covariates. •Identify clinical covariates predicting poor predictive performance among existing popPK models. Methods: PopPK models identified in our prior external evaluation were retained [2-5]. The evaluation cohort comprised of 274 patients providing 2286 concentrations. The previous dataset was further enriched with additional clinically relevant covariates, including residual diuresis, treatment duration, dialyzer model and admission to intensive care unit. Distribution of prediction errors (PE) and absolute PE for population prediction (PRED) were compared according to these covariates, in addition to demographic covariates including age, sex, weight and body mass index (BMI). Logistic regression models using the generalized estimating equations approach were developed to identify covariates associated with overprediction (PE > 20%), underprediction (PE < 20%) and inaccuracy (absolute PE > 30%). Both univariate models (with and without interaction terms for specific popPK models) and a multivariate model were performed, with a significance threshold for p-value set at 0.05. PRED were generated using NONMEM (v7.6; ICON Development Solutions), while distribution plots and logistic regression analyses were performed using the ggplot2 and geepack packages in R (v4.4.2; Posit Software). Results: When comparing the distribution of bias (PE) and imprecision (absolute PE), distinct clusters of increased bias and imprecision were observed in relation to age, treatment duration and BMI. Multivariate logistic regression models indicated that anuria and later stages of the treatment course were associated with both overprediction and inaccuracy. In patients with anuria, the median bias and imprecision were 13.3% and 42.8%, respectively. For patients whose treatment course exceeded two weeks, the median bias and imprecision were 38.6% and 55.6%, respectively. Additionally, higher age, lower BMI and male sex were also linked to overprediction. In patients older than 65 years, those with a BMI below 18.5 kg/m² and in men, the median bias was 20.4%, 17.7%, and 26.8%, respectively. Regression models with interactions showed that the impact of these covariates varied depending on the specific popPK model. Conclusion: These findings suggest that incorporating previously unaccounted-for clinical covariates into future popPK models of vancomycin could help address the variability in patient characteristics within a large multicenter cohort of patients receiving intermittent high-flux hemodialysis. In future analyses, we plan to compare the inclusion of new covariates against other methods, such as model refinement or the development of a meta-model, to produce the optimal popPK model for our population.
Poster Session III, Ιουν?ου 5, 2025, 9:50 πμ – 11:45 πμ 1. Ji C, Garcia J, Sabuga AJ, et al. External evaluation of intravenous vancomycin population pharmacokinetic models in adults receiving high-flux intermittent haemodialysis. Br J Clin Pharmacol. 2025;91(3):856-865. doi:10.1111/bcp.16334 2. Bae SH, Yim DS, Lee H, et al. Application of pharmacometrics in pharmacotherapy: open-source software for vancomycin therapeutic drug management. Pharmaceutics. 2019; 11(5): 224. doi:10.3390/pharmaceutics11050224 3. Goti V, Chaturvedula A, Fossler MJ, Mok S, Jacob JT. Hospitalized patients with and without hemodialysis have markedly different vancomycin pharmacokinetics: a population pharmacokinetic model-based analysis. Ther Drug Monit. 2018; 40(2): 212-221. doi:10.1097/FTD.0000000000000490 4. Hui K, Patel K, Nalder M, et al. Optimizing vancomycin dosage regimens in relation to high-flux haemodialysis. J Antimicrob Chemother. 2019; 74(1): 130-134. doi:10.1093/jac/dky371 5. Oda K, Jono H, Saito H. Model-informed precision dosing of vancomycin in adult patients undergoing hemodialysis. Antimicrob Agents Chemother. 2023; 67(6):e0008923. doi:10.1128/aac.00089-23
Reference: PAGE 33 (2025) Abstr 11593 [www.page-meeting.org/?abstract=11593]
Poster: Clinical Applications