Model-informed precision dosing web tool of busulfan for Brazilian pediatric patients
Laura Ben Olivo (1), Sophia Wermann (1), Bruna Bernar Dias (1), Joice Zuckermann (3), Amanda Valle Pinhati (2,3), Gabriel Giron Correa (2), Liane Esteves Daudt (4), Teresa Cristina Tavares Dalla Costa (1), Bibiana Verlindo de Araújo (1,2)
(1) Pharmaceutical Sciences Graduate Program , Federal University of Rio Grande do Sul, Porto Alegre, Brazil , (2) Medical Sciences Graduate Program, Federal University of Rio Grande do Sul, Porto Alegre, Brazil (3) Pharmacy Service, Hospital de Clínicas de Porto Alegre - HCPA, Federal University of Rio Grande do Sul, Porto Alegre, Brazil (4) Hematology Service, Hospital de Clínicas de Porto Alegre - HCPA, Federal University of Rio Grande do Sul, Porto Alegre, Brazil
Introduction: Therapeutic drug monitoring is widely practiced for busulfan (Bu), with the aim of ensuring exposure into the therapeutic zone [1]. The variability in pharmacokinetics parameters after intravenous administration Bu leads many patients not to achieve the target exposure with standard dosages, especially children [2,3]. Population pharmacokinetic (popPK) modeling emerges as a solution for this focusing on individualized dosage optimization, with the target being the total area under the curve (AUC) around 14,400 - 24,000 uM.min [4]. Several popPK models of Bu for pediatrics were tested in a Brazilian population, however, none of them were able to describe the observed data [5, 6, 7, 8, 9, 10].
Objectives: The aim of this work was to develop and validate a revised popPK model for Bu in Brazilian pediatric patients and apply it to a web useful tool. This work was approved by Ethics Committee # 2.713.246.
Methods: Blood samples were collected at scheduled time points (240, 300, 360 and 420 minutes after Bu 3-hour infusion) for four days. A previous in-house validated UPLC-UV (Ultra Performance Liquid Chromatography) method was used for the quantification of Bu in plasma samples. The data was provided by Hospital de Clínicas de Porto Alegre (HCPA). Data were analyzed based on the population analysis approach using the non-linear mixed effect model with the software NONMEM (version 7.4, ICON Development Solutions, Ellicott City, MD, USA) and PsN version 4.9.0 software (Perl-speaks-NONMEM, Uppsala, Sweden), with the first-order conditional estimation method and interaction (FOCE-I). RStudio (version 2022.02.3), R (version 4.2.1, The R Foundation for Statistical Computing, Vienna, Austria) and pmforest, ggplot2 library as well as Xpose4 library were used for graphical analysis. The interoccasion (IOV) and intraindividual variability (IIV) was analyzed in exponential form. To explain the variability several demographic and biochemical covariates were analyzed. The popPK model was selected according to the minimum value of the objective function (OFV), visual exploration of the goodness of fit plots, and precision of model parameters. A non-parametric bootstrap was performed for the final model (n=1,000) to create the confidence intervals (95%). The final model was externally validated through extra data obtained from HCPA. To evaluate the clinical application of this model, the probability of target attainment (PTA) wwas obtained for the standard protocol, for the model-informed calculated dose, and for the protocols of the tested models [5, 6, 7, 8, 9, 10]. Based on the popPK model developed, an algorithm was created to assess personalized initial doses of Bu. With the algorithm, a web page was developed using HTML5, CSS3 and JavaScript programming language.
Results: A one-compartment model was built from samples of 14 patients (0.5 – 16 y.o.) treated with BU. IIV was explained by adding body weight as a covariate in volume of distribution (Vd) and patients’ age as a covariate in clearance (CL). Both covariates were normalized according to their respective median. The typical value of CL and Vd were estimated to be 3.17 L/h (R.S.E 12%) and 8.61 L (R.S.E 7%), respectively. The IIV for CL was estimated as 39% (R.S.E 16%) and for Vd as 16% (R.S.E 41%). Model validation was performed through a visual predictive check and all bootstrap results stayed within the 95% CI. The algorithm for individualized therapy was Dose (mg)=AUC_target * θ_CL*(1+ θ_AGE*(AGE-43))* e^(ω_CL ). The equation was able to predict target achievement into the therapeutic window in 49.65% of the cases while standard protocol was in 32.05% of the cases, an increase of 17.6 % in success. Our model improves the PTA on target and reduces the probability of first-dose AUC being above/below the target range when compared to other suggested protocols. The final popPK model was implemented as an easy-to-use web page.
Conclusions: In clinical practice, the dosage tool can be used to recommend initial doses of intravenous busulfan based on patient-specific covariate data in a practical and fast way. A specific model for our population provides more secure and precise information that potentially influences positive outcomes of HSCT in pediatric patients. In sum, our results indicate so far that use of this model-based algorithm for precision dose in prospective patients does improve conditioning success.
References:
[1] McCune, J. S. et al. Bone Marrow Transplant. (2002) 30, 167–173
[2] Zao, J. H. et al. Blood Marrow Transplant. (2015) 21, 1471–1478
[3] Marsit, H. et al. Clin. Pharmacokinet. (2020) 59, 1049–1061
[4] McCune, J. S. et al. Clin. Cancer Res. (2014) 20, 754–763
[5] Bartelink IH, et al, Clin Pharmacokinetic. (2012), 51, 331-345.
[6] Booth BP, et al, J Clin Pharmacol. (2007), 47, 101-111.
[7] Long-Boyle JR, et al, Ther Drug Monit. (2015), 37, 236-245.
[8] Nguyen L, et al, Bone Marrow Transplant. (2004), 33, 979-987.
[9] Paci A, et al, Ther Drug Monit. (2012), 34, 198-208.
[10] Poinsignon, V et al, Pediatric blood & cancer (2020), 67, e28603