V Gotta(1), O Marsenic(2), M Pfister(1)
(1) Pediatric Pharmacology & Pharmacometrics, University of Basel Children’s Hospital, Basel, Switzerland (2) Pediatric Nephrology, Yale-New Have Children’s Hospital and Yale School of Medicine, New Haven, CT
Objectives: Delivered urea dialytic clearance (KD) can be predicted from a standard mechanistic equation in adult patients on chronic hemodialysis, based on prescribed blood flow (QB), dialytic flow (QD) and in vitro mass-transfer coefficient for urea of the filter (KoA) [1,2]. It has been suggested that KD is however underpredicted in pediatric hemodialysis patients using this equation [3]. This could be caused by significant differences in dialysis prescription between pediatric and adult patients [3], leading to greater difference (bias) between in vitro determined and in vivo observed KoA in pediatric patients. For example, this may be explained by low and in vitro untested QB and/or QD [4,5]. The aim of this analysis was to improve the prediction of in vivo KoA and thus KD in children on maintenance hemodialysis.
Methods: A two-compartment urea kinetic model previously developed in adults [2] was scaled based on physiologic understanding of kinetic parameters to pediatric patients (total urea distribution volume = total body water [6], distribution clearance scaled by cardiac output [7,8]), with between-subject variability (BSV) assigned to KoA. The model was fitted using NONMEM (FOCE with interaction) to a dataset comprising pre- and post urea concentration samples of 923 pediatric and young adult patients aged 1-29 years (2676 HD sessions in total, with up to 3 hemodialysis sessions (occasions) per patient). Stepwise covariate model building was used to evaluate the association and relationship of prescription parameters with individual in vivo KoA estimates. Nested models were compared by the likelihood ratio test, based on the decrease in objective function value (ΔOFV, α=0.05). The mean relative prediction error (MrPE) in post-HD urea concentrations was further calculated with 95% confidence intervals to assess the bias in population predictions for age groups 1-6, 7-12, 12-18, and 19-29 years.
Results: Increased QD/QB ratio was the parameter most strongly associated with a bias between reported in vitro KoA of the filter and individual estimated in vivo KoA (ΔOFV: -359, p<0.001). A bias of only -4.4% (95% CI: 2.5-6.2%) was estimated at a QD/QB ratio of 2 (reference ratio used in adults), with a linear increase to +114% at a QD/QB ratio of 5 (median ratio used in infants [3], slope estimate: 0.379 (95% CI: 0.34-0.42)). Inclusion of QD/QB ratio as a covariate removed bias observed with QB and QD, and reduced the MrPE in post-HD concentrations largely in children (decrease from 58% (35-81%) to 6% (-9-22%) in children 1-6 years). Inclusion of session duration, filter properties (i.e. low-flux vs high-flux) and ultra-filtration rate could further improve the model fit.
Conclusions: Children on chronic hemodialysis are commonly prescribed much higher QD/QB ratios than adults, which are usually not tested in vitro resulting in the lack of characterization of filter performance for these settings. Although in vitro determined filter KoA is assumed to be a constant, our data shows that this is not the case under in vivo conditions, with largest in vitro-in vivo KoA deviations in youngest patients. Our new model allows prediction of urea dialytic clearance from a range of pediatric specific prescription parameters. The model may also serve to predict drug clearance in pediatric hemodialysis patients.
References:
[1] Michaels et al. Trans Am Soc Artif Intern Organs 1966
[2] Pfister et al. Hemodial. Int. 2008
[3] Gotta et al. (submitted)
[4] Depner et al. ASAIO 2004
[5] Leypoldt et al. ASAIO 2006
[6] Cheek et al. Am. J. Dis. Child 1966
[7] de Simone et al. Circulation 1997
[8] Odeh et al. Clin. Pharmacol. Ther. 1993
Reference: PAGE 27 (2018) Abstr 8724 [www.page-meeting.org/?abstract=8724]
Poster: Drug/Disease Modelling - Paediatrics