CLINICAL TRIALA comparison of group sequential and fixed sample size designsfor bioequivalence trials with highly variable drugsSophie I. E. Knahl 1 & Benjamin Lang 1 & Frank Fleischer 1 & Meinhard Kieser 2Received: 31 July 2017 /Accepted: 9 January 2018 /Published online: 23 January 2018# Springer-Verlag GmbH Germany, part of Springer Nature 2018AbstractPurpose Adrugisdefinedashighlyvariableifitsintra-individualcoefficientofvariation(CV)isgreaterthanorequalto30%.Insuch a case, bioequivalence may be assessed by means of methods that take the (high) variability into account. The ScaledAverage Bioequivalence (SABE) approach is such a procedure and represents the recommendations of FDA. The aim of thisinvestigation is to compare the performance characteristics of classical group sequential designs (GSD) and fixed design settingsfor three-period crossover bioequivalence studies with highly variable drugs, where the SABE procedure is utilized.Methods Monte Carlo simulations were performed to assess type I error rate, power, and average sample size for GSDs withPocock’s and O’Brien-Fleming’s stopping rules and various timings of the interim analysis and for fixed design settings.Results Based on our investigated scenarios, the GSDs show comparable properties with regard to power and type I error rate ascomparedtothecorresponding fixed designs.However, due toanadvantage inaverage samplesize, the mostappealing designisPocock’s approach with interim analysis after 50% information fraction.Conclusions Due to their favorable performance characteristics, two-stage GSDs are an appealing alternative to fixed sampledesigns when assessing bioequivalence in highly variable drugs.Keywords Bioequivalence . Highlyvariabledrugs . Groupsequentialdesigns . Two-stagedesignsIntroductionBioequivalence trials are commonly confirmatory in natureand their aim is to compare the bioavailability of two or moredifferent formulations.In general, bioequivalence is evaluated by constructing a90% confidence interval for the ratio of the geometric means(GMR), which should be completely contained in the pre-specified equivalence interval. This procedure is equivalentto a combination of two one-sided tests, each at a one-sidedalpha level of 5% [1], (U.S. Food and Drug Administration,[2]).Recent scientific research in the field of bioequivalencecomprises investigations that evaluate group sequential andadaptive procedures in classical two-period two-sequencecrossover designs. The possibilities to stop a trial earlier dueto sufficiently meaningful results or to adjust the sample sizeduring the conduct of a trial are attractive properties of thesedesigns. For healthy volunteer trials, this option might be ofparticular interest due to ethical reasons. One of the first dis-cussions with regard to sequential design procedures for bio-equivalence studies was published by Potvin et al. [3] in 2008and further supplemented by Montague et al. [4] and Xu et al.[5]. In 2015, Kieser and Rauch [6] published results of acomprehensive simulation study, which further characterizesthe properties of the procedures that have been proposed byPotvin et al., but also includes general group sequential andadaptive designs, such as Pocock’s design [7, 8] and the in-verse normal method [9].A drug under investigation may be highly variable withrespect to its pharmacokinetic characteristics [10–13]. Suchhighly variable drugs are defined as having an estimatedElectronic supplementary material The online version of this article(https://doi.org/10.1007/s00228-018-2415-7) contains supplementarymaterial, which is available to authorized users.* Meinhard Kiesermeinhard.kieser@imbi.uni-heidelberg.de1Boehringer Ingelheim Pharma GmbH & Co. KG, BirkendorferStraße 65, 88397 Biberach, Germany2Institute of Medical Biometry and Informatics, University ofHeidelberg, Im Neuenheimer Feld 130.3,69120 Heidelberg, GermanyEuropean Journal of Clinical Pharmacology (2018) 74:549–559https://doi.org/10.1007/s00228-018-2415-7
