Weikun Huang1,2, You Zheng1,2, Huiping Huang1,2, Yu Cheng1, Maobai Liu1, Nupur Chaphekar3, Xuemei Wu4,5. 1. Department of Pharmacy, Fujian Medical University Union Hospital, Gulou District, 29 Xinquan Rd., Fuzhou, 350001, Fujian, China. 2. School of Pharmacy, Fujian Medical University, Fuzhou, Fujian, China. 3. Department of Pharmaceutical Sciences, School of Pharmacy, University of Pittsburgh, Pittsburgh, PA, USA. 4. Department of Pharmacy, Fujian Medical University Union Hospital, Gulou District, 29 Xinquan Rd., Fuzhou, 350001, Fujian, China. wuxuemei@fjmu.edu.cn. 5. School of Pharmacy, Fujian Medical University, Fuzhou, Fujian, China. wuxuemei@fjmu.edu.cn.
Abstract
OBJECTIVES: Patients with hematological malignancies are prone to invasive fungal disease due to long-term chemotherapy or radiotherapy. Voriconazole is a second-generation triazole broad-spectrum antibiotic used to prevent or treat invasive fungal infections. Many population pharmacokinetic (pop PK) models have been published for voriconazole, and various diagnostic methods are available to validate the performance of these pop PK models. However, most of the published models have not been strictly evaluated externally. The purpose of this study is to evaluate these models externally and assess their predictive capabilities. METHODS: The external dataset consists of adults receiving voriconazole treatment at Fujian Medical University Union Hospital. We re-established the published models based on their final estimated values in the literature and used our external dataset for initial screening. Each model was evaluated based on the following outcomes: prediction-based diagnostics, prediction- and variability-corrected visual predictive check (pvcVPC), normalized prediction distribution errors (NPDE), and Bayesian simulation results with one to two prior observations. RESULTS: A total of 237 samples from 166 patients were collected as an external dataset. After screening, six candidate models suitable for the external dataset were finally obtained for comparison. Among the models, none demonstrated excellent predictive performance. Bayesian simulation shows that all models' prediction precision and accuracy were significantly improved when one or two prior concentrations were given. CONCLUSIONS: The published pop PK models of voriconazole have significant differences in prediction performance, and none of the models could perfectly predict the concentrations of voriconazole for our data. Therefore, extensive evaluation should precede the adoption of any model in clinical practice.
OBJECTIVES: Patients with hematological malignancies are prone to invasive fungal disease due to long-term chemotherapy or radiotherapy. Voriconazole is a second-generation triazole broad-spectrum antibiotic used to prevent or treat invasive fungal infections. Many population pharmacokinetic (pop PK) models have been published for voriconazole, and various diagnostic methods are available to validate the performance of these pop PK models. However, most of the published models have not been strictly evaluated externally. The purpose of this study is to evaluate these models externally and assess their predictive capabilities. METHODS: The external dataset consists of adults receiving voriconazole treatment at Fujian Medical University Union Hospital. We re-established the published models based on their final estimated values in the literature and used our external dataset for initial screening. Each model was evaluated based on the following outcomes: prediction-based diagnostics, prediction- and variability-corrected visual predictive check (pvcVPC), normalized prediction distribution errors (NPDE), and Bayesian simulation results with one to two prior observations. RESULTS: A total of 237 samples from 166 patients were collected as an external dataset. After screening, six candidate models suitable for the external dataset were finally obtained for comparison. Among the models, none demonstrated excellent predictive performance. Bayesian simulation shows that all models' prediction precision and accuracy were significantly improved when one or two prior concentrations were given. CONCLUSIONS: The published pop PK models of voriconazole have significant differences in prediction performance, and none of the models could perfectly predict the concentrations of voriconazole for our data. Therefore, extensive evaluation should precede the adoption of any model in clinical practice.
Authors: A J Ullmann; J M Aguado; S Arikan-Akdagli; D W Denning; A H Groll; K Lagrou; C Lass-Flörl; R E Lewis; P Munoz; P E Verweij; A Warris; F Ader; M Akova; M C Arendrup; R A Barnes; C Beigelman-Aubry; S Blot; E Bouza; R J M Brüggemann; D Buchheidt; J Cadranel; E Castagnola; A Chakrabarti; M Cuenca-Estrella; G Dimopoulos; J Fortun; J-P Gangneux; J Garbino; W J Heinz; R Herbrecht; C P Heussel; C C Kibbler; N Klimko; B J Kullberg; C Lange; T Lehrnbecher; J Löffler; O Lortholary; J Maertens; O Marchetti; J F Meis; L Pagano; P Ribaud; M Richardson; E Roilides; M Ruhnke; M Sanguinetti; D C Sheppard; J Sinkó; A Skiada; M J G T Vehreschild; C Viscoli; O A Cornely Journal: Clin Microbiol Infect Date: 2018-03-12 Impact factor: 8.067
Authors: Thomas F Patterson; George R Thompson; David W Denning; Jay A Fishman; Susan Hadley; Raoul Herbrecht; Dimitrios P Kontoyiannis; Kieren A Marr; Vicki A Morrison; M Hong Nguyen; Brahm H Segal; William J Steinbach; David A Stevens; Thomas J Walsh; John R Wingard; Jo-Anne H Young; John E Bennett Journal: Clin Infect Dis Date: 2016-06-29 Impact factor: 9.079