Giovanni Smania (1,2), Paola Baiardi (3), Massimo Cella (1,4), Paolo Magni (2)
(1) Consorzio per Valutazioni Biologiche e Farmacologiche, Via Luigi Porta 14, Pavia, Italy, (2) Dipartimento di Ingegneria Industriale e dell’Informazione, Università degli Studi di Pavia, Via Ferrata 5, Pavia, Italy. (3) Direzione Scientifica Centrale, Fondazione Salvatore Maugeri, IRCCS, Via Salvatore Maugeri 4, Pavia, Italy. (4) Pharmatherapeutics Research Clinical Pharmacology, Pfizer Neusentis, Cambridge, UK.
Objectives: The adoption of alternative designs can potentially facilitate the realization of paediatric trials, especially when compared to the standard practice. The objective of this analysis was to compare the performance of the classical parallel design (PaD) with that of a sequential design (based on the Triangular Test (TT) [1]) and of a Bayesian design (BD), exploiting priors from adult data [2], by using PK-PD based clinical trial simulation (CTS).
Methods: A published paediatric PK-PD model of topiramate [3] was used as a paradigm for CTS in epileptic children. 2000 virtual patients were generated based on weight-for-age tables from WHO [4]. Then, for each virtual child, 500 PK profiles were created by using NONMEM version 7.2.0 [5]. Finally, the PD measurement was simulated once for each patient and a design-specific statistical analysis was performed. This procedure was reiterated 1000 times. R software [6] (version 3.0.1) was used for the generation of virtual patients population, drug effect simulations and statistical analyses. Study designs were evaluated in terms of: type I and type II errors, sample size per arm (SS), trial duration (TD) and precision of treatment difference estimate, assessed through the width of its confidence interval (CIW).
Results: Type I and II errors are close to their predetermined levels of 5% and 20% for all designs except for the BD, whose type I error lies between 18.3% and 23.3%. SS for the PaD and BD was fixed to 121 and 54 respectively, whilst the 50th and 95th percentile of sample sizes distribution in the TT design were 80 and 140, respectively. As to TD, assuming an average enrolment rate of 10 patients/month, the PaD and BD showed a TD of 26.2 and 12.8 months, respectively. In contrast, TT’s median TD and its 95th percentile were 32 and 60 months, respectively. Finally, The PaD guarantees the best precision BD and median TT precision are only slightly lower than PaD’s one, even though the value of the 97.5th percentile of CIW distribution in the TT suggests this design could lead to very imprecise estimates.
Conclusions: This model-based analysis suggests that if it is reasonable to assume that treatment effect in children is similar to that in adults, a BD allows to streamline paediatric trials. On the contrary, if adult data cannot be leveraged, a TT design appears to outperform a classical PaD, especially if a large effect size is expected.
This work was supported by the GRiP network (www.grip-network.org).
References:
[1] Whitehead J. The design and analysis of sequential clinical trials. Barnett V, editor. Chichester: Wiley; 1997.
[2] Schoenfeld DA, Hui Zheng, Finkelstein DM. Bayesian design using adult data to augment pediatric trials. Clin Trials (2009); 6(4):297-304
[3] Girgis IG, Nandy P, Nye JS, Ford L, Mohanty S, Wang S, Ochalski S, Eerdekens M, Cox E. Pharmacokinetic-pharmacodynamic assessment of topiramate dosing regimens for children with epilepsy 2 to <10 years of age. Epilepsia (2010); 51(10):1954-62
[4] World Health Organization. “Child growth standards”. http://www.who.int/childgrowth/standards/w eight_for_age/ en/index.html
[5] Beal S, Sheiner LB, Boeckmann A, Bauer RJ, “NONMEM User’s Guides”. (1989-2009), Icon Development Solutions, Ellicott City, MD, USA, 2009.
[6] R Core Team, “R: A Language and Environment for Statistical Computing”, R Foundation for Statistical Computing, Vienna, Austria, 2013, http://www.r-project.org/.
Reference: PAGE 24 (2015) Abstr 3509 [www.page-meeting.org/?abstract=3509]
Poster: Methodology - Study Design