Influence of pointestimates and studypower of bioequivalencestudiesonestablishingbioequivalencebetweengenericsbyadjustedindirectcomparisons

EuropeanJournal of ClinicalPharmacology

Luther Gwaza1, 2, John Gordon3, Henrike Potthast4, Jan Welink5, Hubert Leufkens1, Matthias Stahl6, Alfredo García-Arieta7 *

1 Utrecht Institute for Pharmaceutical Sciences (UIPS), Utrecht, The Netherlands

2 Evaluations and Registration Division, Medicines Control Authority of Zimbabwe, Harare, Zimbabwe

3 Division of Biopharmaceutics Evaluation, Bureau of Pharmaceutical Sciences, Therapeutic Products Directorate, Health Canada, Ottawa, Canada

4 Sub department of Biostatistics and Pharmacokinetics, Federal Institute for Drugs and Medical Devices, Bonn, Germany

5 Medicines Evaluation Board, Utrecht, The Netherlands

6 Prequalification Team – Medicines, Regulation of Medicines and other Health Technologies, Essential Medicines and Health Products, World Health Organization, Geneva, Switzerland

7 División de Farmacología y Evaluación Clínica, Departamento de Medicamentos de Uso Humano, Agencia Española de Medicamentos y Productos Sanitarios, Madrid, Spain

Thismanuscriptrepresentsthe personal opinion of theauthors and doesnotnecessarilyrepresenttheviewsorpolicy of theircorrespondingRegulatory Agencies

*Correspondence:Alfredo García-Arieta

E-mail:

Online resource 1: Brief description of the method to calculate the 90% CI for the adjusted indirect comparisons for each computed scenario.

Statistical analysis of indirect estimates

The CI of the indirect comparison between generics is calculated as: , where is the difference between point estimates of the BE studies, which were defined for each computed scenario (0-14%), t,df is the Student t-value that corresponds to the alpha of 10% in order to estimate CI with 90% confidence and the degrees of freedom of the combination of the BE studies.

The degrees of freedom were calculated as nA+nB-2 in those scenarios where the variability of both studies was the same, whereas in the case of scenarios with different variability they were calculated according to Welch [11], , where nA and nB are the number of subjects in study A for generic A and in study B for generic B, respectively, SDA and SDB are the standard deviations in each BE study (A and B).

The calculation of the SE(d) was performed without any assumption as SEd2=(SEA2+SEB2) in the case of heteroscedasticity and it was calculated as , in the case of homoscedasticity. The combined standard deviation (SDpooled) is derived from the equation , where the degrees of freedom are (nA + nB - 2). The standard deviation (SD) of each individual study is calculated by , where n1 and n2 represents the number of subjects in sequence 1 and sequence 2, respectively for the crossover BE studies.

The standard error (SE(d)) of each study was calculated according to , where n is the number of subjects in the BE study and CV is the coefficient of variation, which was fixed at 25% after verifying that the results are not altered since any change in variability is compensated by the change in sample size required to keep the desired statistical power.