Franziska Schaedeli Stark (1), Clarisse Chavanne (1), Sian Lennon-Chrimes (2), Cheikh Diack (1), Michael Derks (2)
Roche Pharmaceutical Research and Early Development, Pharmaceutical Sciences, (1) Roche Innovation Center Basel, Switzerland; (2) Roche Innovation Center Welwyn, UK
Objectives: Balovaptan is a vasopressin receptor 1a antagonist initially developed to help improve socialization and communication in autism spectrum disorder (ASD). The first study conducted in paediatric participants was a Phase II, randomized, double-blind, 24-week, placebo-controlled study to investigate the efficacy, safety and pharmacokinetics (PK) of balovaptan in children and adolescents with ASD aged 5-17 years (aV1ation [NCT02901431]). We present the model-informed approach used to support the initial selection and the confirmation of age-adjusted doses in aV1ation to achieve a homogenous, adult-equivalent exposure distribution in paediatric participants aged ≥5 years.
Methods: Selection of initial doses in aV1ation, equivalent to 4 mg and 10 mg in adults, was guided by a PBPK model-based approach [2]. Doses were adjusted according to the PBPK-predicted fractions of apparent clearance (CL/F) relative to CL/F in adults, which decreased with age from 95% in older adolescents to 30% in younger children. Four paediatric age/dose bands were proposed, two for adolescents (17-15 and 14-12 yrs), and two for children (11-8 and 7-5 yrs), with initial doses of 4, 3, 2, and 1.5 mg for the 4 mg equivalent arm, and 10, 7, 5, and 3 mg for the 10 mg equvalent arm. The study design included stepwise, age-based recruitment, first in adolescents, then in children, each with an initial cohort of 24 participants (16 on active) for dose confirmation. Once doses were confirmed, full recruitment for the respective age group was opened to the main study part at the confirmed doses. Bayesian forecasting using a population PK (PopPK) model in adults and emerging pediatric week 2 PK data was used to obtain empirical Bayes estimates (EBEs) of paediatric individual PK parameters and AUC at steady-state (AUCss). Due to dose- and time-dependent PK in adults, CL/F was derived from AUCss for adequate comparison with the PBPK model predictions. Doses were confirmed if the median CL/F per age band was within the 90% CI of PBPK model-predicted median values. Otherwise, individual EBEs were used for simulating updated dosing scenarios to achieve adult-equivalent exposures.
Results: Observed CL/F was within the predicted 90% CI for older (17-15 yrs) and younger adolescents (14-13 yrs), this was considered acceptable, and the study was continued at the initial doses. The main study part was opened to adolescents, and the initial cohort in children (12-5 yrs) was opened for dose confirmation. As emerging PK data from the younger participants became available, the overall trend of age-dependent decrease in CL/F appeared weaker than predicted. Based on cumulative data the median CL/F relative to adults decreased from 97% in older adolescents to 59% in younger children. Simulations of exposure supported an updated, simplified dosing algorithm with adult doses for participants aged 8 years and above, and 70% of adult doses for children aged 5-7 yrs, as final doses to be used in the main study part of aViation. Based on cumulative pediatric PK data in aViation, the age effect on CL/F was estimated as a continuous relationship in an updated PopPK model to support the selection of adult-equivalent doses for a study in younger children (aged 2-4 yrs), and for simulations to inform decisions for future pediatric dosing.
Conclusions: A model-informed learn-confirm approach successfully supported the selection and confirmation of paediatric age-adjusted doses for balovaptan. The mismatch between PBPK model-predicted and observed age effect on CL/F, although still within 2-fold, was unexpected, and possibly due to unexplained dose- and time-dependent PK behavior. The study design with narrow initial age bands and staggered, age-based inclusion from older to younger participants appeared complex and sensitive to slow recruitment, but generated informative data for adequate dose confirmation during the study, and for subsequent characterization of pediatric PK in an updated PopPK model.
References:
[1] Bolognani F et al. Sci Trans Med 2019, 11:491, Article Number eaat7838
[2] Schaedeli Stark F et al. PAGE 24 (2015) Abstr 3553 [www.page-meeting.org/?abstract=3553]
Text for reference 1.
Reference: PAGE 30 (2022) Abstr 9988 [www.page-meeting.org/?abstract=9988]
Poster: Drug/Disease Modelling - Paediatrics