Standards for model-based early bactericidal activity analysis and sample size determination in tuberculosis drug development
Laurynas Mockeliunas (1), Alan Faraj (1), Rob C van Wijk (1), Caryn M. Upton (2), Gerben van den Hoogen (2), Andreas H. Diacon (2), Ulrika S.H. Simonsson (1)
(1) Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Sweden, (2) TASK, Cape Town, South Africa
Introduction/Objectives: One of the milestones in tuberculosis (TB) drug development is a phase 2a trial, also known as an early bactericidal activity (EBA) study. The purpose of the EBA study is to identify if a drug or treatment shows activity in TB patients during a short period up to 14 days. Despite EBA trials being conducted for many years, the amount of information about the best practices for EBA trial analysis is limited, including the estimation of the sample size required to avoid underpowering. This work aimed to present a standardized pharmacometric model-based EBA analysis workflow and characterize sample sizes needed to detect EBA or the difference between two treatment arms.
Methods: Seven steps were identified for a standardized pharmacometric model-based EBA analysis approach: data exploratory analysis, base model development, covariate model development, EBA detection, pharmacokinetic-pharmacodynamic (PK-PD) modeling, EBA comparison, and EBA reporting. A non-linear mixed-effects modeling approach was used to exemplify the process of using the standardized approach for EBA analysis, as well as to perform sample size calculations using the Monte-Carlo Mapped Power method to detect EBA and to detect a difference between two arms. A mono-exponential time-to-positivity-EBA (TTP-EBA) simulation model composed of parameters from two previously developed TTP-EBA models for two meropenem-containing treatments [1-2] was used to generate data. Two replicates per each time point per participant were simulated. TTP observations were simulated for days 0, 1, 2, 3, 4, 5, 8, 10, 12 and 14. To characterize the sample sizes needed to detect EBA, TTP slope values corresponding to TTP-EBA0-14 ranging from 3 hours to 152 hours, as well as inter-individual variability (IIV) in TTP slope of 22% (coefficient of variation [CV]) and 104% CV were used. To characterize the sample size needed to detect a difference between the two treatment arms, TTP slope values corresponding to TTP-EBA0-14 of 30 hours and 152 hours, IIV in TTP slope ranging from 10% CV to 104% CV, and effect difference between two treatments ranging from 25% to 200% were used. Target power was 80% at a 5% significance level.
Results: This standardized pharmacometric model-based EBA analysis approach is composed of seven steps allowing to derive EBA through the estimation of TTP slope, IIV in TTP slope, incorporation of statistically significant covariates as well as pharmacokinetics. With this approach, EBA detection, treatment comparison, and even dose optimization can be performed for current and new biomarkers. Based on sample size determinations for EBA detection, 13 and 8 participants/arm were sufficient to detect TTP-EBA0-14 of 11 hours when IIV in TTP slope was 104% CV and 22% CV, but higher sample sizes were required for lower EBA. According to sample size determinations to detect an effect difference between two arms, the smallest sample sizes were required when TTP-EBA0-14 of 152 hours and IIV in TTP slope of 22% CV were present. For this scenario, 21 participant/arm was sufficient to detect an effect difference of 25%, and less than 10 participants/arm were required for higher effect differences. Higher sample sizes were required to detect the effect difference for TTP-EBA0-14 of 30 hours and IIV in the TTP slope of 22% CV scenario. In this case, to detect an effect difference of 25%, 64 participants/arm were required, while a lower number of participants was required for higher effect differences. The presence of high IIV in TTP slope required larger sample sizes compared to low IIV in TTP slope scenarios. For high IIV in TTP slope scenarios (104% CV), large sample sizes were required to detect an effect difference between two arms, and 15 participants/arm would not be sufficient to detect a difference between two arms in most of the explored cases, except when TTP-EBA0-14 of 30 hours was present and the effect difference was at least 175%.
Conclusions: This work presented a standardized pharmacometric model-based EBA analysis approach which was established in close collaboration between pharmacometricians, microbiologists, and clinicians. Additionally, sample size determinations were performed, which provided a better understanding of factors influential on the power needed to detect EBA or find a difference between two treatment arms.
References:
[1] De Jager, V., Gupte, N., Nunes, S., Barnes, G. L., van Wijk, R. C., Mostert, J., et al. (2022). Early Bactericidal Activity of Meropenem plus Clavulanate (with or without Rifampin) for Tuberculosis: The COMRADE Randomized, Phase 2A Clinical Trial. Am J Respir Crit Care Med 205, 1228–1235. doi: 10.1164/rccm.202108-1976OC.
[2] Unpublished data, ClinicalTrials.gov Identifier: NCT04629378
