Model-based approach for group sequential and adaptive designs in parallel and cross-over bioequivalence studies
Manel Rakez (1), Julie Bertrand (1), Florence Loingeville (1), Thu Thuy Nguyen (1), France Mentré (1), Andrew Babiskin (2), Guoying Sun (3), Stella Grosser (3), Liang Zhao (2) and Lanyan (Lucy) Fang (2)
(1) INSERM, IAME, UMR 1137, F-75018 Paris, France, (2) Office of Research and Standards, Office of Generic Drugs, Center for Drug Evaluation and Research, Food and Drug Administration, 10903 New Hampshire Ave, Silver Spring, MD 20993, USA, (3) Office of Biostatistics, Office of Translational Sciences, Center for Drug Evaluation and Research, Food and Drug Administration, 10903 New Hampshire Ave, Silver Spring, MD 20993, USA
Introduction: Bioequivalence (BE) studies are performed to compare pharmacokinetics (PK) of drug formulations. Traditionally, the two one-sided test (TOST) is performed using estimates of AUC and Cmax obtained by non-compartmental analysis (NCA). Assumptions on the expected variability of AUC and Cmax are needed for sample size calculation. In case of uncertainty, it has been recently proposed to perform two-stage studies considering group sequential and adaptive designs [1]. In a previous work [2], we proposed a model-based TOST as an alternative to NCA-based TOST.
Objectives:To extend model-based statistical approaches for BE assessment to two-stage group sequential and adaptive designs and evaluate them by clinical trial simulation.
Methods: Group sequential design, with fixed number of subjects at each stage, and adaptive design, with interim sample size re-estimation, were developed for model-based TOST using the Pocock method [3] and the standard combination test [1], respectively. We evaluated parallel and cross-over rich sampling PK designs. We generated, under H0 and H1 hypothesis, 500 simulated data sets for each scenario and design. The parameters of the nonlinear mixed effect models were estimated using Monolix software and asymptotic standard errors were used. Total sample size, type I error and power were compared between single-stage, group sequential and adaptive two-stage designs.
Results: We implemented, in R, two-stage parallel and cross-over BE designs using model-based approach with tests on the treatment effect. The expected Population Fisher Information Matrix was used to compute the number of subjects needed at the second stage of adaptive designs [4]. Clinical trial simulations illustrated the good properties of the approach with preserved type I errors. Total sample sizes were generally reduced when two-stage design was performed with no loss of power compared to single-stage design. The benefit was even more striking with adaptive design when a higher variability was assumed at the planning stage than the one actually obtained from the first stage.
Conclusions: We showed that BE assessment using two-stage designs is feasible using model-based approach, which is an extension of these designs for NCA BE [1]. Further extensions are needed for sparse design where asymptotic standard errors can be too small [2].
Acknowledgements: This work was supported by the Food and Drug Administration (FDA) under contract 10110C. The authors thanks FDA for this funding. The views in this abstract do not necessarily reflect the views or policies of the FDA.
References:
[1] Maurer W, Jones B, Chen Y. Controlling the type I error rate in two-stage sequential adaptive designs when testing for average bioequivalence. Stat Med. 2018; 37(10):1587-1607.
[2] Dubois A, Lavielle M, Gsteiger S, Piegeolet E, Mentré F. Model-based analyses of bioequivalence crossover trials using the SAEM algorithm. Stat Med. 2011;30:2582-2600.
[3] Kieser M, Rauch G. Two-stage designs for cross-over bioequivalence trials. Stat Med. 2015; 34(16):2403-16.
[4] Dumont C, Lestini G, Le Nagard H, Mentré F, Comets E, Nguyen TT and the PFIM group. PFIM 4.0, an extended R program for design evaluation and optimization in nonlinear mixed-effect models. Comput Methods Programs Biomed. 2018;156:217–29.
