Ligelizumab Paediatric Investigation Plan: exposure-response analysis in adult chronic spontaneous urticaria with simulation-based design of adolescent dose-finding
Philip J. Lowe (1), Stephan Köhne-Voss (1), Richard Mills (2), Colm Farrell (2)
(1) Novartis Pharma AG, Basel, Switzerland, (2) ICON plc, Dublin, Ireland.
Objectives: Ligelizumab is an investigational monoclonal antibody that binds immunoglobulin E (IgE) with higher affinity than omalizumab and has the potential to benefit a greater proportion of patients with chronic spontaneous urticaria (CSU). In establishing the paediatric investigational plan (PIP) due note was made of a significantly reduced potency, i.e. higher EC50 of omalizumab in the adolescent CSU population [1] compared with adults. It could not therefore be assumed that equivalent concentrations of ligelizumab in adolescent and adults would result in equivalent efficacy in adult and adolescent populations. The objective was to design an adequate adolescent study to determine whether the EC50 for adolescent CSU patients was sufficiently different from adults as to demand a different posology.
Methods: Ligelizumab concentrations and urticaria activity scores (UAS7, 7 day sum of daily itch and hives, each with range 0-3) were collected from adult patients treated with placebo, low, medium and high dose levels every 4 weeks multiple and a high single dose ligelizumab in a Phase 2 study [2]. Interim data from 295 patients were analysed with longitudinal pharmacokinetic-pharmacodynamic models for the continuous UAS7 (range 0-42) using NONMEM with importance sampling. The resultant model was used with stochastic simulation-estimation to design an adolescent (age 11-17 years inclusive) study with the ability to detect shifts in EC50 between adolescents and adults. A combination of R-3.2.3, NONMEM 7.3.0 [3] and PDx-Pop-5.2 software was used to create analysis datasets, estimate parameters, control the NONMEM runs and post-process results.
Results: The chosen two-compartment pharmacokinetic model described well the drug concentration data. The key exposure parameter, clearance, was 0.85 L/d (residual standard error, RSE, 9.1%) for 80 kg bodyweight with 49% coefficient of between subject variation (BSV). Bodyweight was identified as the main covariate impacting clearance with an estimate of 1.0 (power; 35% RSE). The chosen continuous UAS7 model had an EC50 of 1.1 µg/mL (38% RSE) with very large estimated BSV (1405%) and a steep Hill coefficient of 5.72 (0.75% RSE). Visual prediction checks were deemed sufficient to initiate the adolescent study design process over a number of study options. To maintain numerical stability the BSV on EC50 was reduced from that estimated to <=300%, with a sensitivity analysis included to investigate the impact. The design chosen specified three arms: placebo, low and high dose levels every 4 weeks with a treatment duration of 16 weeks and follow-up to 40 weeks. The placebo patients should cross to the high dose after 8 weeks.
Conclusions: Despite highly variable data the exposure-UAS7 response model was able to detect and describe reasonably well placebo and ligelizumab dose-related changes over time. Estimates of ligelizumab clearance and EC50 potency for reducing the signs and symptoms of urticaria were as expectated from previous clinical studies and analyses thereon [4,5,6]. Stochastic simulation-estimation indicated that a design with two active dose levels plus a placebo control should suffice for the prospective adolescent study. The low dose was prioritised as this would generate concentrations in the region of the expected EC50, the optimum point of sensitivity for estimating this parameter. The randomisation was therefore uneven, with 20 patients on the low dose, 10 patients each on the high dose and placebo. The high dose, from both the directly treated and crossed-over placebo patients, would enable estimation of the maximum drug effect. Overall, based on 100 simulation-estimations, the procedure indicated that there was approximately 80% chance to detect a 2-fold increase in EC50, the threshold above which a different posology from adults should be considered. The pharmacometric analysis will be the subject of a separate pooled modelling and simulation study as per PIP.
References:
[1] EMA 2014. Application II/0048 Assessment report, Section 2.3.4.2 page 10 and Section 2.3.5 page 11. http://www.ema.europa.eu/docs/en_GB/document_library/EPAR_-_Assessment_Report_-_Variation/human/000606/WC500164453.pdf accessed 17th January 2018.
[2] EudraCT 2014-005559-16. http://www.clinicaltrialsregister.eu/ctr-search/trial/2014-005559-16/GB accessed 17th January 2018.
[3] Bauer RJ NONMEM Users Guide Introduction to NONMEM 7.3.0. 2013. Icon Development Solutions, Hanover MD 21076, USA.
[4] Lowe PJ, Fink M, Milton MN. Two case studies on how study designs can be made more informative using modeling and simulation approaches. Clin Pharmacol Ther 2017; 102 (6): 908-911.
[5] Gauvreau GM, Arm JP, Boulet LP, Leigh R, Cockcroft DW, Davis BE, Mayers I, FitzGerald JM, Dahlen B, Killian KJ, Laviolette M, Carlsten C, Lazarinis N, Watson RM, Milot J, Swystun V, Bowen M, Hui L, Lantz A-S, Meiser K, Maahs S, Lowe PJ, Skerjanec A, Drollmann A, O'Byrne PM. Efficacy and safety of multiple doses of QGE031 (ligelizumab) versus omalizumab and placebo in inhibiting allergen-induced early asthmatic responses. J Allergy Clin Immunol 2016; 138(4): 1051-9.
[6] Lowe PJ, Kümmel A, Vasalou C, Matsushima S, Skerjanec A, Integrated quantitation of biotherapeutic drug–target binding, biomarkers, and clinical response to support rational dose regimen selection. In: ADME and Translational Pharmacokinetics/Pharmacodynamics of Therapeutic Proteins: Applications in Drug Discovery and Development, First Edition, Zhou H, Theil FP, editors. Chapter 13. John Wiley & Sons, Inc. 2016.
