Stefano Grosdani 1, Pietro Laddomada 1, Oscar Della Pasqua 1
1 Clinical Pharmacology & Therapeutics Group, University College London (London, United Kingdom)
Introduction
MYCN is a key oncogenic driver in several paediatric malignancies, including neuroblastoma, where its amplification is associated with rapid disease progression, treatment resistance, and poor prognosis. BGA002 is a first-in-class anti-gene peptide nucleic acid designed to selectively silence MYCN at the DNA level, resulting in tumour growth inhibition in preclinical models[1]. Nonclinical studies demonstrated significant antitumour activity with sustained tumour exposure[2]. However, substantial inter-individual variability has been observed across preclinical species, making the identification of the therapeutic dose range in humans a major challenge[3,4].
Given this translational uncertainty, together with the inherent complexity of conducting a first-in-human (FIH) trial in paediatric patients with a rare disease, an adaptive, model-informed design has been proposed, where dynamic dose adaptations are based on observed systemic exposure (AUC). Three cohorts are expected to receive 28-day intravenous doses targeting predefined BGA002 AUC levels of 350, 700, and 1,400 h·ng/mL.
An exposure-guided dosing algorithm will be used to individualize treatment. Patients will initially receive a priming dose, derived from extrapolated preclinical disposition parameters, to achieve a predefined target AUC. During the first two treatment days, six 1mL blood samples per day will be collected for interim pharmacokinetic (PK) analysis, to build an individual PK profile. Individual disposition parameters will then be estimated and used to guide dose adjustment.
Objectives
The aim of this work is to illustrate the implementation of a model-guided FIH Phase I/II adaptive protocol to evaluate the safety, tolerability, pharmacokinetics and preliminary efficacy of intravenous BGA002 in paediatric patients with refractory MYCN-altered tumours, enabling optimized exposure-based dosing bands, while addressing uncertainties in drug disposition and probability of pharmacological success.
Methods
Clinical trial simulations were implemented using a previously developed population pharmacokinetic model. Simulation scenarios were evaluated considering a target exposure range, defined previously through the integration of in vitro and in vivo data, describing MYC-N amplification inhibition, cell viability, and reduction in tumour weight in a xenograft model[1].
Simulations were based on an initial protocol which aims to include a total of 18 paediatric patients aged 2 to 18 years, divided into two cohorts. Here, we focus on methodological challenges related to dose selection in cohort 1. Given disease rarity, patients are expected to be enrolled one at a time. Baseline characteristics of interest included body weight (range 10 kg – 60 kg; expected median approximately 40 kg. Three clinical scenarios were simulated to reflect the uncertainty in the extrapolation of disposition properties of BGA002:
1. Scenario 1: target exposure is achieved as predicted by allometric scaling principles.
2. Scenario 2: lower systemic exposure, with deviation ≥50% of the intended target.
3. Scenario 3: higher systemic exposure (i.e., ≥150%), consistent with findings in another nonclinical toxicology species.
Sampling time optimisation and parameter estimation were implemented using informative and non-informative priors in conjunction with a simulation-re-estimation procedure [5]. Clearance estimates were based on allometric scaling, assuming a 40 kg patient. Recommendations for dose adjustment were made under the assumption of dose-proportional, time-invariant pharmacokinetic properties. All modelling and simulation steps were implemented in NONMEM 7.5.1 and PsN 5.3.1. Data handling, graphical and statistical summaries were performed in R 4.1.2.
Results
Individual clearance and systemic exposure were adequately estimated using semi-informative priors. Predicted median (5th–95th percentile) clearance estimates ranged between 4.4 (3.3–6.2), 30.5 (21.23–41.9), and 58.4 (45.3–83.2) L/h across the different scenarios. Predicted median AUC values varied between were 149.9 (116.1–195.3), 300.1 (222.5–389.9) and 2183.6 (1661.7–2835.9) ng·h/mL, ensuring low probability of exposure above the predefined safety threshold for a single dose.
An optimised sampling schedule, including blood sampling before and 1.0, 1.5, 2.0, 3.0, 6.0, 24, 25, 26.5, 28, 36 and 48 after the first dose, enabled accurate estimation of pharmacokinetic parameters, confirming the appropriateness of the priming doses. This schedule was also adequate in scenario 3, where significantly lower clearance and prolonged elimination half-life were simulated. Achievement of target AUCs was confirmed by subsequent pharmacokinetic sampling of dose adjusted regimens on day 28.
Conclusions
This model-guided FIH Phase I/II adaptive protocol provides a robust framework for dose individualisation in a paediatric oncology setting. The selected PK sampling schedule was sufficient to generate reliable individual exposure estimates, enabling dose optimisation and, thus, mitigation of translational uncertainty.
References:
[1] A. L. Scardovi et al., “Preclinical Pharmacokinetics in Tumors and Normal Tissues of the Antigene PNA Oligonucleotide MYCN-Inhibitor BGA002,” Nucleic Acid Ther., vol. 34, no. 4, pp. 173–187, Aug. 2024, doi: 10.1089/nat.2024.0005.
[2] S. Oosterholt, A. L. Scardovi, R. Tonelli, and O. Della Pasqua, “PKPD modelling of MYCN-inhibition in vitro and in vivo in a mouse model of neuroblastoma,” 2017.
[3] Lin R, Yin G, Shi H. Bayesian adaptive model selection design for optimal biological dose finding in phase I/II clinical trials. Biostatistics. 2023; 24(2):277-294.
[4] Qiu Y, Li M. A Bayesian dynamic model-based adaptive design for oncology dose optimization in phase I/II clinical trials. Pharm Stat. 2025; 24(2):e2451.
[5] S. C. Van Dijkman et al., “Dose rationale and pharmacokinetics of dexmedetomidine in mechanically ventilated new-borns: impact of design optimisation”, doi: 10.1007/s00228-019-02708-y.
Reference: PAGE 34 (2026) Abstr 12212 [www.page-meeting.org/?abstract=12212]
Poster: Methodology - Study Design